Understanding the Requirements of a Decision Support System for Integrated Production in Agriculture Anna Perini SRA Division ITC-irst Via Sommarive 18, 38050 Povo (TN) - Italy http://sra //sra.itc.it i* Symposium April 20th 2005 - City University, London Outline The Problem Domain Understanding the Domain» Main questions» Domain analysis: Actor / goal modeling The Decision Support System» Main requirements» Changes in domain organizations Conclusion 2 1
The Integrated Production Domain Integrated Production in Agriculture consists of a set of practices aimed at favoring the set up of a development model characterized by a reduced environmental impact. Plant Disease management, according to IP, consists in using lowimpact and/or natural techniques to maintain the disease damage of a crop under a specific tolerance threshold. Developing DSS at support of IP application requires to take into account two main dimensions of complexity: the organizational dimension dealing with all the dependencies among domain stakeholders involved in management tasks (e.g. the producers, the technicians of the advisory service, government agencies); the technical dimension concerning the modeling of environmental phenomena. 3 The Apple Codling Moth (apple bug) case a critical pest for apple it is resistant to pesticides specialized chemicals for eggs /larvae/ adults pest management requires a high number of actions in a season Eggs Larvae Adults The CM life cycle in Trentino: at least 2 generations critical events: the starting of eggdrop qualitative models for predicting the pest evolution tend to fail due to the area s heterogeneous orography 5/1/2002 5/15/2002 5/29/2002 6/12/2002 6/26/2002 7/10/2002 7/24/2002 8/7/2002 8/21/2002 4 2
Understanding the Domain Main questions Who performs the management activities? Why? Are there any organizational processes these activities rest on? or Who depends on whom for what? What are the critical info/knowledge? Which are most the critical, decisional steps in these processes? Are there alternative ways to perform a decision making steps? Which decision could be supported by a DSS? Who will be the user? Analysis approach Interviews of producers, technicians and domain experts, acquisition of domain documentation. Analysis of actor roles and of the strategic dependencies between actors, goal-analysis, plan-analysis. 5 Actor / goal modeling Domain stakeholders profit Work. Healthy Env ironm ent Producer reg istration tradem ark follow IP protocol follow EU rules Local Government favour IP production Plant Disease Expert Advisor Geographic distribution of actors needs to be taken into account. 6 3
IP domain strategic dependencies The Local Government funds an Advisory Service to support the producers in using IP practices and research on novel agronomic techniques Where does the strategic dependencies between Advisor and Producer come from? What is the role of the Plant Disease Expert in the strategic dependency network? 7 IP domain strategic dependencies cntd 8 4
Management of the Apple Codling Moth Before eggdrops I generation II generation Preventive actions Using Pheromone Trapping systems to lower mating Periodic assessment of the infection degree and choice of a remedy action (disease management plan) What are the data and knowledge these management activities rest on? And who owns them? geometrical features of the orchards and information on the presence of infection sources, diffusion barriers are needed to design a PT system models to predict the disease evolution on the basis of a limited set of environmental data will support the definition of more effective disease management plans 9 The Advisor goals 10 5
IP domain strategic dependencies Focusing on the CM case The Producer delegates to the Advisor the goal of designing an appropriate PT system for her/his orchard. The Advisor depends on the Producer to get the necessary data on the orchard. 11 The DSS for the Management of the Apple CM Before eggdrops I generation II generation Preventive actions Using Pheromone Trapping systems to lower mating The advisor uses orchards maps and takes into account geometrical features and information on the presence of infection sources, diffusion barriers, etc. to design a PT system Periodic assessment of the infection degree and choice of a remedy action (disease management plan) On the basis of a regular collection of data, the expert defines models to predict the disease evolution, the advisor identifies alerting events that will be communicated to the producers The DSS users and tasks 12 6
Designing a Pheromone Trapping system Analyze production data and identify infection sources 680 679 680 679 678 567/1 678 563 567/1 567/2 561/1 561/2 563 567/2 561/1 560 561/2 560 Production info Ton 561/1 Melo m9 Ton 561/2 Melo m9 Get geographical and land registry s maps Report on Pheromone usage on the area Town Particella rootstock #dispenser Ton 561/1 m9 200 13 GIS based design of a PT system Area managed by the advisor Production info for this area Choice of the orchards where PT will be installed Report of PT features for the area 14 7
How does it affect the IP strategic dependencies? DSS It will acquire data on behalf of the advisor (the user) It will make orchard data directly available to the advisors. These data are collected by producers organizations in a DB accessed by the DSS It will support also novice technicians to do a good job 15 Collecting and analyzing seasonal data Organizing observations in a spredsheet Annotating Monitoring egg drop in the field Observ ations In 2003 Gathering meteo data Data integration Analysis Meteo data 16 8
Collecting and analyzing seasonal data Automate Organizing observations in a spredsheet Extend the approach Annotating Monitoring egg drop in the field Observ ations In 2003 Gathering meteo data Data integration Analysis Meteo data 17 The DSS supporting data collection and analysis for disease management plan definition Annotating Client Data integration Server Export in Excel Meteo and disease data per area Anaysis Monitoring egg drop in the field Analysis with ML techniques 18 9
How does it affect the IP strategic dependencies? DSS It will acquire data on behalf of the advisor (the user) It will make orchard data directly available to the advisors. These data are collected by producers organizations in a DB accessed by the DSS It will support also novice technicians to do a good job It will provide the advisor with new disease models and made available to the experts data produced by the advisors. 19 Conclusion An agent-oriented software development methodology which provides intentional analysis techniques has been used for: understanding the strategic interests of domain stakeholders and their mutual dependencies for goal achievement; analyzing the system requirements and the architecture. Lessons learned: An organization model can be effective in supporting the communication between the analysts and the domain stakeholders (including users) Allow to model (and reason on!) process reengineering 20 10
Acknowledgment and References What has been presented is based on a joint work with Angelo Susi (ITC-irst, SRA Division) The dss-pico project has been partially funded by the Italian Ministry for University and Research (MIUR). Partners: IASM-du, IASM-cat, APOT, PAT, ITC-irst. Papers Perini, Anna; Susi, Angelo, "Supporting Decision Making in Plant Disease Management", Published in Proceedings of the 4th Workshop on Binding Environmental Sciences and Artificial Intelligence [BESAI 2004], pp. 3-1/3-7. Ref. No. P04-10-03 A. Perini and A. Susi. Designing a Decision Support System for Integrated Production in Agriculture. An Agent-Oriented approach. Environmental Modelling and Software Journal, 19(9), September 2004. 21 11